close

Вход

Забыли?

вход по аккаунту

?

j.eneco.2017.10.015

код для вставкиСкачать
Accepted Manuscript
Market Deregulation and Nuclear Safety
Zhen Lei, Chen-Hao Tsai
PII:
DOI:
Reference:
S0140-9883(17)30356-0
doi:10.1016/j.eneco.2017.10.015
ENEECO 3788
To appear in:
Received date:
Revised date:
Accepted date:
17 April 2017
13 October 2017
16 October 2017
Please cite this article as: Zhen Lei, Chen-Hao Tsai , Market Deregulation and Nuclear
Safety. The address for the corresponding author was captured as affiliation for all authors.
Please check if appropriate. Eneeco(2017), doi:10.1016/j.eneco.2017.10.015
This is a PDF file of an unedited manuscript that has been accepted for publication. As
a service to our customers we are providing this early version of the manuscript. The
manuscript will undergo copyediting, typesetting, and review of the resulting proof before
it is published in its final form. Please note that during the production process errors may
be discovered which could affect the content, and all legal disclaimers that apply to the
journal pertain.
ACCEPTED MANUSCRIPT
Market Deregulation and Nuclear Safety
Zhen Lei* and Chen-Hao Tsai**
RI
PT
Abstract
Nuclear reactor initiating events, which are unplanned reactor emergency shutdowns, are a
primary measure of reactor safety performance. We call attention to several methodological and
SC
data issues in the emerging research endeavor that uses nuclear initiating events to investigate the
impacts of electricity market deregulation on nuclear reactors’ safety performance. Correcting
NU
these issues we find that the effects of plant divestiture on nuclear safety are much smaller in
MA
magnitude and less significant. Moreover, we find that when examining data prior to 2008 or
excluding years 2008 and 2009 from analysis, the effect of plant divestiture on reactor initiating
D
events becomes robustly insignificant.
PT
E
Keywords: Deregulation, Nuclear safety, Reactor initiating events
JEL codes: D22, L51, L94
CE
Acknowledgements: We are grateful for helpful comments from Andrew N. Kleit, the Editor,
Richard Tol, and three anonymous referees. All errors remain our own.
AC
* John and Willie Leone Family Department of Energy and Mineral Engineering and EMS
Energy Institute, 110 Hosler Building, Pennsylvania State University, University Park, PA
16802. (Email: zlei@psu.edu; Tel: 814-863-0810; Fax: 814-865-3248)
** Corresponding author. Bureau of Economic Geology’s Center for Energy Economics,
Jackson School of Geosciences, The University of Texas at Austin, Austin TX 78758 (Email:
chenhao.tsai@beg.utexas.edu; Tel: 713-654-5403; Fax: 713-654-5405)
ACCEPTED MANUSCRIPT
Highlights:
Nuclear reactor initiating events are a primary measure of safety performance
•
Researches use initiating events to study impact of deregulation on nuclear safety
•
We discuss two methodological issues in extant studies
•
We propose a better data source on nuclear initiating events
•
We find that the effects of deregulation on nuclear safety are inconclusive
AC
CE
PT
E
D
MA
NU
SC
RI
PT
•
ACCEPTED MANUSCRIPT
1.
Introduction
Modern nuclear power reactors are designed and engineered to achieve a very low risk of
serious accidents. Nonetheless, events such as 1979 Three Mile Island and 2011 Fukushima
accidents have sustained a long-standing concern about the safety of nuclear power plants. In the
RI
PT
United States, electricity market restructuring since mid-1990s has brought an additional layer of
uncertainty, as approximately half of the commercial nuclear reactors in the U.S. have been
SC
deregulated. The U.S. Nuclear Regulatory Commission (NRC) has been raising the concern
about “the possible effects that rate deregulation and disaggregation…could have on nuclear
NU
plant operational safety.”1 Economic theories lead to ambiguous predictions on how
deregulation impacts nuclear plants’ performance in safety. On one hand, one could argue that
MA
deregulation incentivizes a nuclear plant to cut corners in its pursuit for short-term returns,
resulting in a decrease in safety performance; on the other hand, safety might be a complement to
D
operating performance and thus emphasizing safety might be good business for nuclear
PT
E
generators. Therefore, understanding the impacts of deregulation on nuclear safety performance
is clearly an important empirical question.
CE
There has been an emerging line of research aims at gathering rigorous empirical evidence
for this key question, including intriguing studies by both Davis and Wolfram (2012) and
AC
Hausman (2014). These studies used reactor initiating events, which are unplanned reactor
emergency shutdowns (a.k.a. “trips” or “scrams”), as a primary measure of nuclear plants safety
performance.2 Their results show that the frequency of reactor initiating events has decreased
following plant divestiture.3
1
See the NRC Final Policy Statement (62 FR 44071), dated August 19, 1997.
2
Hausman (2014) also uses other safety measures including fires, escalated enforcement actions,
collective worker radiation exposure, and average worker radiation. However, only the results based on
reactor initiating events are robust and consistently suggest that nuclear reactors’ safety performance
Page 1
ACCEPTED MANUSCRIPT
In this paper we call attention to two potentially important methodological issues in this
research area. We also point to another dataset on initiating events that has only recently become
publically available, which we believe to be a better data source than that used by the previous
studies. Addressing these issues in part or altogether, we show that the effect of plant divestiture
RI
PT
on the frequency of reactor initiating events becomes smaller than those reported in Davis and
Wolfram (2012) and Hausman (2014).
SC
We further vary the study period in our analyses. We find that if we examine data prior to
2008 or excluding years 2008 and 2009 from analyses, the effect of plant divestiture on the
NU
frequency of reactor initiating events becomes robustly insignificant, and much smaller in
magnitude.
MA
Our study highlights a more careful use of initiating events as an indicator for nuclear
operational safety, and points to the need for more empirical evidence on the effects of electricity
D
market restructuring on nuclear safety. The remainder of the paper is organized as follows. In
PT
E
Section 2 we discuss two conceptual and methodological issues in the literature. In Section 3 we
introduce a better data source for reactor initiating events. We provide the empirical results in
CE
Section 4 and conclude in Section 5.
AC
2. Two Conceptual and Methodological Issues
Issue #1: Initiating events with “plant-centered” versus “external” root causes
A reactor initiating event involves an unplanned reactor trip, either automatically or manually,
that a reactor licensee or operator is obliged to report to the NRC in a Licensee Event Report
improves following plant divestiture.
3
In both studies, divestiture involves either regulated utilities transfer their nuclear plants to an
unregulated subsidiary, or sale to an independent power producer.
Page 2
ACCEPTED MANUSCRIPT
(LER).4 Since initiating events are likely to “upset plant stability and challenge critical safety
functions” (U.S. NRC, 1999), they are one of the key safety measures used in the NRC reactor
oversight programs.5 In particular, the NRC establishes three performance indicators based on
initiating events, including (i) Unplanned Scrams per 7000 Critical Hours; (ii) Unplanned Power
RI
PT
Changes per 7000 Critical Hours; and (iii) Unplanned Scrams with Complications, to track
reactor safety performance on a quarterly basis. The NRC also maintains and updates
periodically a database of initiating events, and performs analyses on initiating events from
SC
probabilistic risk perspectives.6
NU
The extant studies use all initiating events reported by a nuclear reactor as the measurement
of changes in its internal safety performance before and after plant divestiture. However,
MA
initiating events can be triggered due to either “plant-centered” root causes that reflect internal
performance of plant operator, or by “external” root causes that do not. For example, Fermi
D
nuclear plant in Michigan experienced a scram on November 15, 2007, in which the root cause
PT
E
was “a failure to adhere to procedures and a less than adequate pre-job brief.” Apparently this
initiating event was related to the performance of plant personnel and thus is of a “plant-centered”
CE
root cause. In contrast, the same plant had another automatic emergency shutdown during the
August 2003 North America Blackout. For this event, the root cause was a large scale grid
4
AC
disturbance that was an “external” root cause, which the nuclear plant operator had no control.7
Pursuant to Title 10 Code of Federal Regulations Part 50 Section 73 (10 CFR 50.73) – Licensee Event
Report System, a reactor licensee is obligated to submit an report to the NRC within 60 days after the
discovery of a reportable event defined under §50.73(a)(2).
5
See NRC Detailed Reactor Oversight Process Description:
http://www.nrc.gov/reactors/operating/oversight/rop-description.html, and NRC Inspection Procedures &
Performance Indicators by Reactor Oversight Process Cornerstone:
http://www.nrc.gov/NRR/OVERSIGHT/ASSESS/cornerstone.html
6
See NRC Reactor Operational Experience Results and Databases: http://nrcoe.inel.gov/resultsdb/
7
See LER 341-2007-002 (dated January 14, 2008) and LER 341-2003-002 (dated December 17, 2003)
for the details about these two initiating events.
Page 3
ACCEPTED MANUSCRIPT
More specifically, “plant-centered” root causes are incidents that occur within the physical
boundary of nuclear plants, including (1) equipment or component failures; and (2) plant
personnel operation errors or management deficiency. On the other hand, “external” root causes
are those beyond the reach of nuclear plant operators, including (1) faults or disturbance over the
RI
PT
transmission grid;8 (2) natural phenomenon;9 and (3) interference of foreign material, wide
animals, or aquatic life.10
SC
It seems reasonable that, for the purpose of evaluating the impact of deregulation on nuclear
plants’ internal safety performance, one should focus on initiating events with “plant-centered”
NU
root causes. Events with “external” root causes are often beyond the control of the plant operator
and thus not a function of plants’ internal behavior.
MA
As an evidence, the NRC requires that the plant owners describe in LERs what “corrective
actions” they can implement to make the plants more resilient in the aftermath of an initiating
D
event. For “plant-centered” events plant owners often report specific corrective actions;11 for
PT
E
events with “external” causes, however, plant owners tend to report corrective actions that often
fall outside the responsibility of nuclear plants or only state that “potential long-term corrective
8
CE
actions are being evaluated.”12 Hence it’s unclear whether plant owners could take measures that
AC
For instance, the Unit 2 of the Indian Point nuclear plant (New York) experienced reactor trips twice
within a period of four months in 2003, due to repeated failure of nearby transmission lines that are
owned and maintained by Consolidates Edison. See LER 247-2003-003 (dated June 27, 2003) and LER
247-2003-004 (dated October 2, 2003).
9
For example, during a severe thunderstorm on July 23, 1995, at least one direct lightning stroke Vogtle
plant and triggered reactor trip signals. See LER 424-1995-002 (dated August 18, 1995).
10
Foreign interference can be unusually large build-up of fish or jellyfish near circulating water intakes
(see, e.g. LER 335-1993-007, and LER 313-1998-005).
11
For events caused by equipment failures, corrective actions often include immediate repair of damaged
or malfunctioned equipment, which often involve long-term improvement in equipment design, material,
or manufacture. For events caused by human errors, corrective actions often include better training,
communication and management practice, or modification of procedures.
12
For instance, Indian Point nuclear plant experienced multiple emergency shutdown in 2003 because of
repeated failure of nearby transmission line owned and maintained by Consolidated Edison. Entergy, the
plant owner of Indian Point, reported to NRC that their corrective actions were to follow-up with
Page 4
ACCEPTED MANUSCRIPT
would render nuclear plants more resilient to external causes.
One respectful referee argues that managerial efforts could make a difference in mitigating
the impact on plant operation of events with “external” causes and therefore those events may
still be relevant for evaluating the safety performance of a nuclear plant. As shown in Section 4,
RI
PT
when including all events (regardless root causes), we still find a weaker correlation between
plant divestiture and safety performance than reported in Hausman (2014). See Table A1 in the
SC
Appendix. Thus our results are not driven by excluding events with “external” causes.
NU
Issue #2: Normalization of the frequency of initiating events
As pointed out by the extant literature, to measure a reactor’s safety performance it is
MA
important to use normalized annual frequency of initiating events that controls for the level of
operating time in a year. Studies (Zhang, 2007; Davis and Wolfram, 2012) have shown that,
D
following deregulation and plant divestiture, a reactor increases operating time and generates
PT
E
more electricity. With more operating time in a year, a deregulated reactor is more likely to
experience initiating events even if its safety performance remains unchanged.13 Hence, to
CE
examine the effects on plant safety performance, it is a reactor’s annual operating time, not its
electricity generation output, that needs to be controlled for: if a reactor’s generation capacity
AC
increases, the same level of generation output does not mean the same operating time.
Thus, how to do the normalization is also an important methodological issue and worth
careful deliberation. Hausman (2014) uses capacity factor, the ratio of the reactor’s electricity
Consolidated Edison, including meeting between senior levels of management of two companies (see
LER 247-2003-003). In other “external” events involving interference of foreign materials or aquatic life,
plant owners often make vague statements such as “potential long-term corrective actions are being
evaluated.”
13
Indeed, Hausman (2014) finds that plant divestiture has no significant effects on un-normalized
frequency of initiating events.
Page 5
ACCEPTED MANUSCRIPT
generation in a year to the original design generation capacity (which is time-invariant), as a
proxy of operating time for normalizing the frequency of initiating events. We argue that this
normalization might be problematic because a reactor’s original design capacity could be smaller
than the actual generation capacity in a given year: many reactors in the U.S. have since
RI
PT
mid-1990s performed nuclear power uprates to increase their generation capacity beyond the
original licensed limit. Thus, using the original design capacity for calculating capacity factor
SC
might overstate a nuclear reactor’s operating time. With an overstated operating time, the
normalized frequency of initiating events would be understated. Moreover, deregulation tends to
NU
incentivize nuclear reactors to invest in power uprates (Davis and Wolfram, 2012; Lei et al.,
2017). Thus normalizing the frequency of initiating events using capacity factors could lead to a
MA
spurious negative correlation between deregulation and normalized frequency of initiating events
(and safety performance).
D
We also argue that a “correct” capacity factor, the ratio of the reactor’s electricity generation
PT
E
to the current generation capacity that takes into account power uprates, might still be a noisy
proxy of plant operating time. The EIA data on electricity generation is the net generation by a
CE
plant, which is the gross generation subtracting the plant load (electricity generation consumed
by the plant itself, which amounts to 3% to 5% of the gross generation and varies from one plant
AC
to another).
Here, we propose to normalize the annual frequency of initiating events by reactor critical
year, the percentage of time that a reactor is under normal operating condition in a calendar year.
Reactor critical year is a direct measure of a plant’s operating time and availability, bounded
between the value of zero and one and not affected by nuclear power uprates or electricity
consumed by plant itself. Indeed, the NRC monitors reactor operating status and see if a reactor
Page 6
ACCEPTED MANUSCRIPT
deviates from “critical” (i.e. 100% power) daily.14 The NRC also uses “Unplanned Scrams per
7,000 Critical Hours” and “Unplanned Power Changes per 7,000 Critical Hours” to track reactor
safety performance on a quarterly basis.15
To illustrate the difference between reactor critical year and capacity factor, consider the R.E.
RI
PT
Ginna nuclear plant (in New York) that had a 16.8% power uprate in 2006. The plant, which ran
nearly full time and at the full capacity in 2007, recorded a 99% critical year and a 113%
SC
capacity factor for that year. In comparison, Diablo Canyon nuclear plant unit 2 (in California),
which also ran full time at full capacity in 2007 but had no power uprates, recorded a 100%
NU
critical year and 99% capacity factor.
MA
3. A Better Data Source on Initiating Events
The extant studies use initiating events data tables in the NRC report “Rates of Initiating
D
Events at U.S. Nuclear Power Plants 1988-2010”, as the primary data source for initiating
PT
E
events.16 Since these data tables are summary tables compiled by the NRC for the purpose of
engineering risk assessment, their use needs careful treatment. First, the owner of a nuclear plant
CE
may report two separate initiating events in one LER if they are close in timing, in which case
the NRC assigns one LER number to both events. Since initiating events data tables are compiled
AC
based on LER numbers, only one of the events, but not both, are included in the data tables. Thus,
using these complied data tables would neglect some initiating events. For instance, the Arkansas
One nuclear plant reported two separate initiating events (one on Dec. 12, 2008 and the other on
14
See Power Reactor Status Reports
https://www.nrc.gov/reading-rm/doc-collections/event-status/reactor-status/
15
Data of “Operating Information Critical Years (1980-2015)” is retrieved on November 7, 2016, from
http://nrcoe.inl.gov/resultsdb/ReactorYears/
16
The current and historical NRC initiating events reports and the data tables are available
at http://nrcoe.inel.gov/resultsdb/InitEvent/
Page 7
ACCEPTED MANUSCRIPT
Dec. 20, 2008) in the same licensee event report (LER 313-2008-001). One of them is not
included in the data tables compiled by the NRC and thus omitted in Hausman (2014). Indeed,
there are about 70 initiating events between 1988 and 2009 that are omitted in Hausman
(2014).17
RI
PT
Second, and more importantly, since nuclear power plants are designed and engineered to act
in a chainlike fashion upon an incidence, an initiating event might involve initial plant faults
SC
(IPFs) and some subsequent functional impacts (FIs), in which case the NRC records the event
based on both the IPFs and subsequent FIs. Accordingly, an initiating event could be listed in
NU
more than one data table in the NRC report. Thus, simple data extraction from initiating events
data tables could lead to multiple counting of an initiating event. Take for an example the
MA
initiating event that occurred to Calvert Cliffs (Maryland) Unit 1 on March 20, 2004 (LER
317-2004-001). The event was originated from a feedwater regulating valve shut, leading to
D
rapid lowering of the cooling water level in the steam generator (because of the loss of feed
PT
E
water flow). In turn, the steam generator feed water pumps tripped due to a high discharge
pressure, resulting in a loss of normal heat removal and eventually an automatic reactor trip. This
CE
single event was listed by the NRC in two different data tables, in Table 22 (Loss of Feedwater
LERs) and Table 31 (PWR loss of heat sink LERs) in the NRC report “Rates of Initiating Events
AC
at U.S. Nuclear Power Plants 1988–2010.” Thus it is counted twice in Hausman (2014).
We propose to use the NRC spreadsheet file on initiating events, which the NRC has since
2012 made publically available, as the data source.18 The spreadsheet file records the
information including the IPFs and FIs for each initiating event reported to the NRC, and the
NRC uses it as the basis for its report “Rates of Initiating Events at U.S. Nuclear Power Plants”
17
For due diligence, we went through original LER filings to ensure that there were indeed multiple
initiating events reported in the same LER.
18
The latest spreadsheet file of initiating events is available at http://nrcoe.inel.gov/resultsdb/InitEvent/
Page 8
ACCEPTED MANUSCRIPT
and the data tables in the report. Using this NRC spreadsheet file averts the problems of omitting
or multiple counting initiating events.
4. Empirical Results
RI
PT
We use the most recent NRC spreadsheet file on initiating events as the data source, which
includes initiating events reported to the NRC during the period between 1988 and 2014. We
identify the root causes for each initiating event in the data, by manually reviewing its LER
SC
where nuclear operators are required to specify and report the root causes.19,20 Excluding events
NU
with “external” or “indeterminate” root causes,21 we construct a data on the frequency of
initiating events with “plant-centered” root causes, at the reactor-year level.
MA
If we compare the data compiled in Hausman (2014) that covers the period of 1988-2009 to
our data for the same period,22 counting initiating events in the NRC spreadsheet file (to fix data
D
omission or duplication errors) reduces the total number of events from 2,817 to 2,612, and the
PT
E
average number of events per reactor-year from 1.25 to 1.16. When we focus on “plant-centered”
events, the number of events further drops to 2,317 and the average number of events per
CE
reactor-year 1.03. For the period of 1988-2014, the total number of initiating events counted in
the NRC spreadsheet file is 2,901 and the average number of events per reactor-year is 1.05,
19
AC
The requirement of reporting root causes and the designation of root causes are pursuant to NRC
NUREG-1022 “Event Report Guidelines: 10 CFR 50.72 and 50.73,” which was published in 1983 and
has since been consistent (there were three revisions, in 1998, 2000, and 2013, respectively, aiming at
clarifying and resolving the ambiguities in the guidelines). The NRC reviews each submitted event report
and requests plant owners to made necessary modification/revision should NRC considers that a report
fails to meet the criteria specified in the NUREG-1022 guidelines.
20
We first reviewed the summary of an initiating event’s LER to identify the root causes. If root causes
were not mentioned or unclear in the LER summary, we then looked into the LER’s detailed description.
LERs are accessible at https://lersearch.inl.gov/Entry.aspx.
21
About 9.7% of the initiating events in the data are with “external” root causes (see Table B1 in Online
Appendix). There are a few initiating events for which the root causes were reported as indeterminate in
the LERs. These events are excluded from our analyses (see Table B2 in Online Appendix for the list).
22
The data in Hausman (2014) is downloadable from
https://www.aeaweb.org/articles.php?doi=10.1257/pol.6.3.178.
Page 9
ACCEPTED MANUSCRIPT
while the total number of “plant-centered” events is 2,572 and the average number is 0.93. See
Table 1.
Next, we use our data and run the same four econometric models as in Hausman (2014), to
estimate the impacts of plant divestiture on reactor safety performance. The regressions involve
RI
PT
regressing the number of initiating events on a divestiture dummy and a set of reactor fixed
effects and year effects. The four models include negative binomial, conditional negative
SC
binomial, Poisson and linear panel data regressions, given that the dependent variable, the
frequency of initiating events in a year for a reactor, is a count data variable.
NU
Table 2 exhibits the results with step-by-step corrections. In Panel A, we replicate and show
the results in Hausman (2014). In Panel B, we use the frequency of initiating events, including
MA
both with “plant-centered” causes and with external causes, which are counted in the NRC
spreadsheet file (to avert the problems of omitting or multiple counting of events). We also keep
D
the same study period (1988-2009) and the same normalizing variable (capacity factor) as in
PT
E
Hausman (2014).23 In Panel C, we focus on the frequency of “plant-centered” initiating events
that are counted in the NRC spreadsheet file, while maintaining the study period of 1988-2009
CE
and capacity factor as the normalizing variable. The results indicate that, relative to the results in
Hausman (2014), the magnitude (absolute value) of the estimated coefficients for
AC
drops in these four regressions.
In Panel D of Table 2, we further change the normalizing variable in the regressions, using
reactor critical year rather than capacity factor. The magnitude (absolute value) of the estimated
23
Capacity factors are calculated based on the same data source EIA Form 923 and the same design
capacity used by Hausman (2014). In Hausman (2014), normalization for the count regressions is
accomplished by including capacity factor as an independent variable with the coefficient on the logged
variable equal to one; normalization for linear panel regressions is by dividing the dependent variable by
capacity factor. Hausman (2014) also drops observations with capacity factor less than 0.01, which was
also replicated here.
Page 10
ACCEPTED MANUSCRIPT
coefficients for
further drops in all four regressions, and in one regression (linear
panel) the coefficient for
becomes less than one half than reported in Hausman
(2014) and turns insignificant. Finally, in Panel E, we expand the study period to 1988-2014. The
decreases further and the coefficients
become either insignificant or significant only at 10% level.
RI
PT
magnitude of the estimated coefficients for
The further decrease in the magnitude of the estimated coefficients when expanding the
SC
study period led us to examine the robustness of the results by varying the study period, as
shown in Table 3. In all the regressions, we focus on “plant-centered” events counted in the NRC
NU
spreadsheet file and use reactor critical year for normalization. Using the data prior to year 2008,
the estimated coefficient for
is small in magnitude and insignificant (Panel A and
MA
B). When data in years 2008 and 2009 is included in the analyses, the coefficient for
increases in absolute value, and becomes statistically significant (Panel C and D)
D
in three of the four regressions. When including data after 2009, the coefficient for
PT
E
turns smaller in absolute value and less significant statistically (Panel E through panel I).
It appears that including years 2008 and 2009 in the analyses increases the absolute value
CE
and significance of the estimated coefficient for
. Thus, in Panel J of Table 3, we
examine the study period of 1988-2014 but exclude years 2008 and 2009. The analysis yields
AC
small and insignificant estimator for
across all four regressions, suggesting that
indeed it is years 2008 and 2009 that give rise to significant results.
Therefore, there are two interesting findings from this analysis. First, the impact of plant
divestiture on nuclear plant safety performance is inconclusive and mostly likely to be
insignificant if excluding years 2008 and 2009 from analyses. Second, deregulated reactors
seemed to have better performance in nuclear safety than regulated ones in 2008 and 2009 when
Page 11
ACCEPTED MANUSCRIPT
the financial crisis unfolded. Indeed, for regulated reactors the number of “plant-centered”
initiating events per reactor-year was 0.53 in 2006-2007 and increased to 0.81 in 2008-2009 (the
difference is statistically significant with p value=0.0146), whereas for divested reactors the
figure dropped from 0.60 in 2006-2007 to 0.38 in 2008-2009 (p value= 0.0402).24 We carefully
RI
PT
study all initiating events reported in years 2006-2009, but could not identify consistent
engineering causes that might be attributable to these changes during the period. It is an
SC
interesting research question that merits further investigations in the future.
In Appendix A, we further conduct several robustness checks. We first check if the results in
NU
Table 3 hold when we made step-by-step changes. In Table A1, we counts all initiating events in
the NRC spreadsheet file, including both with “plant-centered” causes and with external causes,
MA
while keeping capacity factor as the normalizing variable (as in Hausman, 2014). In Table A2,
we focus on “plant-centered” initiating events counted in the NRC spreadsheet file, while
D
maintaining capacity factor as the normalizing variable. The results indicate that the estimated
are statistically insignificant when examining data prior to 2008 or
PT
E
coefficients for
excluding years 2008 and 2009.
CE
There are four nuclear plants closed permanently between 2013 and 2014, either due to
technical issue or financial consideration.25 In Table A3 of Appendix A, we exclude these four
remain similar to the main results.
AC
plants in analyses. The coefficient estimates of
In another robustness check, we aim to separate the potential impacts of deregulation on
24
Using the data in Hausman (2014), for regulated reactors the numbers of initiating events per
reactor-year were 0.63 in 2006-2007 and 0.8 in 2008-2009 (p value= 0.1899), and for divested reactors
they were 0.73 and 0.38 for the two periods respectively (p value =0.0084).
25
These nuclear plants include San Onofre (CA), Crystal River (FL), Kewaunee (WI) and Vermont
Yankee (VT). San Onofre (CA) and Crystal River (FL) were closed because of the cost and uncertainty of
repairing plant equipment or structure, and there were extended outages for about one to two years
preceding their permanent closures. Kewaunee (WI) and Vermont Yankee (VT) on the other hand were
closed due to economic consideration, and they were under normal operation until the day of closure.
Page 12
ACCEPTED MANUSCRIPT
nuclear safety through installing major power uprates from those via affecting plant internal safety
performance. Deregulated reactors are found to be more likely to undertake major power uprates
including stretch power uprates (SPU) and extended power uprates (EPU) (Lei et al., 2017), which
in turn could affect nuclear safety. In Table A4 of Appendix A, we included additional independent
RI
PT
variables to account for the possible effects of installing SPUs and EPUs on reactor safety. The
coefficient estimates of
remain similar to the main results reported in Table 3.
SC
Moreover, the impacts of these two major uprates on the frequency of initiating events are
insignificant.
NU
In a final robustness check, we investigate whether the insignificant results for the
prior-to-2008 study periods hold if we use the data in Hausman (2014) and capacity factor as the
MA
normalizing variable. As shown in Table A5 of Appendix A, the estimated coefficients for
D
are smaller in absolute value and turn statistically insignificant in most of cases.
PT
E
5. Conclusions
This paper is motivated by several potential methodological and data issues in the emerging
CE
research that uses nuclear initiating events as a key measure of reactor safety performance and
investigates the effects of electricity market deregulation on nuclear safety. These potential issues
AC
include: (1) distinguishing initiating events with “plant-centered” root causes from those with
“external” root causes; (2) normalization of the annual frequency of initiating events; and (3)
data sources on initiating events. Addressing these issues, our analyses find the effects of plant
divestiture on nuclear plants’ safety performance to be smaller in magnitude and less significant
statistically than those reported in previous studies. Furthermore, when we vary the study period,
Page 13
ACCEPTED MANUSCRIPT
the coefficient for
becomes insignificant for data prior to 2008 or data excluding
years 2008 and 2009.
Hence our paper calls for more empirical evidences on the impacts of electricity market
deregulation on nuclear safety. It might be important for researchers to be mindful of the
AC
CE
PT
E
D
MA
NU
SC
empirical results by, for example, varying the study period.
RI
PT
methodological and data issues discussed in this paper, and be careful in examining robustness of
Page 14
ACCEPTED MANUSCRIPT
REFERENCE
Davis, Lucas W. & Catherine Wolfram (2012). “Deregulation, Consolidation, and Efficiency:
Evidence from U.S. Nuclear Power.” American Economic Journal: Applied Economics, 4(4),
RI
PT
194-225.
Hausman, Catherine (2014). “Corporate Incentives and Nuclear Safety.” American Economic
Journal: Economic Policy, 6(3): 178-206.
SC
Lei, Zhen, Chen-Hao Tsai, and Andrew N. Kleit (2017). “Deregulation and Investment in
Generation Capacity: Evidence from Nuclear Power Uprates in the United States.” The Energy
Journal, 38(3): 113-139.
NU
U.S. NRC (1999). “NUREG/CR-5750 Rates of Initiating Events at U.S. Nuclear Power Plants:
1987-1995.” Retrieved from U.S. NRC website: http://nrcoe.inel.gov/resultsdb/InitEvent/
MA
U.S. NRC (2014). “NUREG-1649 Reactor Oversight Process.” Retrieved from U.S. NRC
website: http://www.nrc.gov/reading-rm/doc-collections/nuregs/staff/sr1649/r5/
Zhang, Fan (2007) “Does Electricity Restructuring Work? Evidence from the U.S. Nuclear
AC
CE
PT
E
D
Energy Industry.” The Journal of Industrial Economics, 55(3), 397.
Page 15
ACCEPTED MANUSCRIPT
Table 1: Data comparison: Initiating events
Number of initiating events, at reactor-year level
Mean
Standard
Deviation
Replication from data in
Hausman (2014)
1.254
1.493
Include All Initiating Events,
counted from NRC spreadsheet
1.163
1.321
Include Plant-Centered
Initiating Events, counted from
the NRC spreadsheet file
1.032
1.249
Include All Initiating Events,
counted from NRC spreadsheet
1.053
1.266
Include Plant-Centered
Initiating Events, counted from
the NRC spreadsheet file
0.933
Min
Total number of
initiating events
Max
10
2,817
0
9
2,612
0
9
2,317
NU
0
9
2,901
1.194
0
9
2,572
SC
0
MA
Panel B: For the period of 1988-2014
RI
PT
Panel A: For the period of 1988-2009
AC
CE
PT
E
D
Note: Our data, as in Hausman (2014), involves 103 investor-owned nuclear reactors operating in the U.S. Browns Ferry Unit
1, which was in prolonged outage during the study period and was restarted in 2005/2006, is excluded. In our data, there are
in total 2,751 and 2,245 reactor-year observations for the periods of 1988-2014 and 1988-2009, respectively (several reactors
started commercial operation after 1988).
ACCEPTED MANUSCRIPT
Table 2: Effects of Plant Divestiture on Normalized Frequency of Initiating Events
Negative Binomial
(1)
Conditional NB
(2)
Poisson
(3)
Linear Panel
(4)
Panel A: Results from Hausman (2014) (which we replicate), for the period of 1988-2009
-0.335***
(0.122)
Divestiture
-0.34***
(0.12)
-0.32**
(0.12)
-0.54**
(0.23)
-0.311***
(0.117)
Divestiture
RI
PT
Panel B: Using all initiating events counted from the NRC spreadsheet file, for the period of 1988-2009
-0.313**
(0.124)
-0.300**
(0.117)
-.0463**
(0.200)
Panel C: Using plant-centered events counted from the NRC spreadsheet file, for the period of 1988-2009
-0.329**
(0.131)
-0.316**
(0.125)
SC
-0.327***
(0.124)
Divestiture
-0.412**
(0.200)
NU
Panel D: Using plant-centered events counted from the NRC spreadsheet file, and normalizing by Critical Year, for
the period of 1988-2009
-0.289**
(0.126)
Divestiture
-0.294**
(0.133)
-0.284**
(0.126)
-0.240
(0.190)
Reactor fixed effects
Year dummies
X
X
D
-0.219*
(0.120)
PT
E
Divestiture
MA
Panel E: Using plant-centered events counted from the NRC spreadsheet file, and normalizing by Critical Year, for
the period of 1988-2014
-0.219*
(0.125)
-0.214*
(0.120)
-0.166
(0.188)
X
X
X
X
X
X
AC
CE
Note: We run unconditional negative binomial, conditional negative binomial, Poisson, and linear panel regressions, as in
Hausman (2014). The dependent variable in all regressions is the number of initiating events for a reactor in a year. In
Panel A, we replicate the regressions using the data from Hausman (2014) that covers the period of 1988-2009. In Panel B,
we change the dependent variable to all initiating events counted in the NRC spreadsheet file (to avert the problems of
omitting or multiple counting events), while maintaining the study period of 1988-2009 and capacity factor as the
normalizing variable. In Panel C, we changed the dependent variable to “plant-centered” initiating events counted in the
NRC spreadsheet file, while maintaining the study period of 1988-2009 and capacity factor as the normalizing variable. In
Panel D, we further change the normalizing variable from capacity factor to reactor critical year, while maintain the study
period of 1988-2009. In Panel E, we expand the study period to 1988-2014 and focus on “plant-centered” initiating events
counted in the NRC spreadsheet file and use reactor critical year as the normalizing variable. Following Hausman (2014),
normalization for the count regressions in Column (1) to (3) is accomplished by including capacity factor or reactor
critical year as an exposure variable (i.e. as an independent variable with the coefficient on the logged variable equal to
one); normalization for linear regressions in Column (4) is by dividing the dependent variable by capacity factor or reactor
critical year. The regressions focus on 103 investor-owned nuclear reactors operating in the U.S. (Browns Ferry Unit 1,
which was in prolonged outage during the study period and was restarted in 2005/2006, is excluded). All regressions
include reactor fixed effects and year dummies. Following Hausman (2014), robust standard errors are clustered at plant
level in Negative Binomial, Poisson, and Panel Linear regressions; for Conditional NB regressions, standard errors are
bootstrapped (and also clustered at plant level) with 500 replications. *** p<0.01, ** p<0.05, * p<0.1
Page 17
ACCEPTED MANUSCRIPT
Table 3: Effects of Plant Divestiture: Varying the Study Period
Using plant-centered events counted from the NRC spread file, and normalizing by Critical Year
Poisson
(3)
Linear Panel
(4)
Panel A: for the period of 1988-2006
-.147
Divestiture
(0.119)
-0.146
(0.123)
-0.142
(0.120)
-0.162
(0.198)
Panel B: for the period of 1988-2007
-.127
Divestiture
(.123)
-0.131
(0.126)
-0.123
(0.123)
-0.139
(0.199)
Panel C: for the period of 1988-2008
-.244**
Divestiture
(0.122)
-0.247*
(0.127)
-0.238*
(0.123)
-0.214
(0.196)
Panel D: for the period of 1988-2009
-0.290**
Divestiture
(0.126)
-0.294**
(0.133)
-0.284**
(0.127)
-0.240
(0.190)
Panel E: for the period of 1988-2010
-0.257**
Divestiture
(0.123)
-0.255*
(0.130)
-0.250**
(0.124)
-0.215
(0.188)
Panel F: for the period of 1988-2011
-0.237**
Divestiture
(0.121)
-0.238*
(0.127)
-0.232*
(0.122)
-0.191
(0.186)
-0.228*
(0.128)
-0.224*
(0.121)
-0.182
(0.187)
-0.232*
(0.128)
-0.230*
(0.122)
-0.180
(0.187)
-0.219*
(0.125)
-0.214*
(0.120)
-0.166
(0.188)
-0.101
(0.121)
-0.0963
(0.192)
CE
Panel I: for the period of 1988-2014
-0.219*
Divestiture
(0.120)
SC
MA
PT
E
Panel H: for the period of 1988-2013
-0.236*
Divestiture
(0.122)
D
Panel G: for the period of 1988-2012
-0.228*
Divestiture
(0.121)
RI
PT
Conditional NB
(2)
NU
Negative Binomial
(1)
AC
Panel J: for the period of 1988-2014 but excluding 2008 and 2009
-0.105
-0.103
Divestiture
(0.121)
(0.125)
Note: We run unconditional negative binomial, conditional negative binomial, Poisson, and linear panel regressions, as in
Hausman (2014). The dependent variable in all regressions is the number of “plant-centered” initiating events counted in
the NRC spreadsheet, by reactor-year. We use reactor critical year as the normalizing variable and vary the study period in
these regressions. Following Hausman (2014), all regressions include reactor fixed effects and year dummies, with robust
standard errors clustered at plant level in Negative Binomial, Poisson, and Panel Linear regressions; for Conditional NB
regressions, standard errors are bootstrapped (and also clustered at plant level) with 500 replications. *** p<0.01, **
p<0.05, * p<0.1
Page 18
ACCEPTED MANUSCRIPT
Online Appendix A:
AC
CE
PT
E
D
MA
NU
SC
RI
PT
Robustness Checks
Page 19
ACCEPTED MANUSCRIPT
Poisson
(3)
Linear Panel
(4)
Panel A: for the period of 1988-2006
-0.180
Divestiture
(0.113)
-0.178
(0.115)
-0.170
(0.113)
-0.395*
(0.206)
Panel B: for the period of 1988-2007
-0.164
Divestiture
(0.114)
-0.167
(0.117)
-0.156
(0.114)
-0.372*
(0.207)
Panel C: for the period of 1988-2008
-0.268**
Divestiture
(0.112)
-0.270**
(0.118)
-0.258**
(0.113)
-0.439**
(0.203)
Panel D: for the period of 1988-2009
-0.311***
Divestiture
(0.117)
-0.313**
(0.124)
Panel E: for the period of 1988-2010
-0.284**
Divestiture
(0.115)
-0.463**
(0.201)
-0.277**
(0.122)
-0.272**
(0.116)
-0.434**
(0.200)
-0.282**
(0.118)
-0.276**
(0.114)
-0.425**
(0.200)
-0.275**
(0.120)
-0.272**
(0.115)
-0.432**
(0.205)
Panel H: for the period of 1988-2013
-0.277**
Divestiture
(0.114)
-0.271**
(0.119)
-0.265**
(0.114)
-0.416**
(0.203)
Panel I: for the period of 1988-2014
-0.255**
Divestiture
(0.114)
-0.250**
(0.119)
-0.244**
(0.114)
-0.394*
(0.203)
-0.143
(0.114)
-0.332
(0.205)
NU
-0.300**
(0.117)
MA
Negative Binomial
(1)
SC
Conditional NB
(2)
RI
PT
Table A1: Effects of Plant Divestiture: Varying the Study Period
Using all initiating events counted from NRC spreadsheet, and normalizing by Capacity Factor
CE
PT
E
Panel G: for the period of 1988-2012
-0.283**
Divestiture
(0.115)
D
Panel F: for the period of 1988-2011
-0.287**
Divestiture
(0.114)
AC
Panel J: for the period of 1988-2014 but exclude 2008 and 2009
-0.152
-0.147
Divestiture
(0.114)
(0.117)
Note: We run unconditional negative binomial, conditional negative binomial, Poisson, and linear panel regressions, as in
Hausman (2014). The dependent variable in all regressions is the number of initiating events counted in the NRC
spreadsheet file, by reactor-year. We use capacity factor as the normalizing variable and vary the study period in these
regressions. Following Hausman (2014), all regressions include reactor fixed effects and year dummies, with robust
standard errors clustered at plant level in Negative Binomial, Poisson, and Panel Linear regressions; for Conditional NB
regressions, standard errors are bootstrapped (and also clustered at plant level) with 500 replications. *** p<0.01, **
p<0.05, * p<0.1
ACCEPTED MANUSCRIPT
Table A2: Effects of Plant Divestiture: Varying the Study Period
Using plant-centered events from the NRC spreadsheet file, and normalizing by Capacity Factor
Conditional NB
(2)
Poisson
(3)
Linear Panel
(4)
Panel A: for the period of 1988-2006
-0.176
Divestiture
(0.118)
-0.172
(0.121)
-0.166
(0.118)
-0.336
(0.207)
Panel B: for the period of 1988-2007
-0.157
Divestiture
(0.122)
-0.160
(0.124)
Panel C: for the period of 1988-2008
-0.277**
Divestiture
(0.120)
-0.279**
(0.125)
Panel D: for the period of 1988-2009
-0.327***
Divestiture
(0.124)
-0.329**
(0.131)
Panel E: for the period of 1988-2010
-0.296**
Divestiture
(0.123)
RI
PT
Negative Binomial
(1)
-0.314
(0.208)
-0.267**
(0.121)
-0.384*
(0.204)
-0.316**
(0.125)
-0.412**
(0.201)
-0.290**
(0.129)
-0.285**
(0.124)
-0.386*
(0.199)
-0.275**
(0.126)
-0.269**
(0.122)
-0.361*
(0.198)
-0.268**
(0.127)
-0.263**
(0.121)
-0.371*
(0.202)
Panel H: for the period of 1988-2013
-0.280**
Divestiture
(0.123)
-0.272**
(0.128)
-0.270**
(0.123)
-0.365*
(0.202)
Panel I: for the period of 1988-2014
-0.264**
Divestiture
(0.121)
-0.260**
(0.126)
-0.255**
(0.121)
-0.350*
(0.201)
-0.139
(0.122)
-0.284
(0.204)
AC
CE
NU
MA
PT
E
D
Panel F: for the period of 1988-2011
-0.278**
Divestiture
(0.121)
Panel G: for the period of 1988-2012
-0.272**
Divestiture
(0.122)
SC
-0.149
(0.122)
Panel J: for the period of 1988-2014 but excluding 2008 and 2009
-0.147
-0.143
Divestiture
(0.123)
(0.125)
Note: We run unconditional negative binomial, conditional negative binomial, Poisson, and linear panel regressions, as in
Hausman (2014). The dependent variable in all regressions is the number of “plant-centered” initiating events counted in
the NRC spreadsheet file, by reactor-year. We use capacity factor as the normalizing variable and vary the study period in
these regressions. Following Hausman (2014), all regressions include reactor fixed effects and year dummies, with robust
standard errors clustered at plant level in Negative Binomial, Poisson, and Panel Linear regressions; for Conditional NB
regressions, standard errors are bootstrapped (and also clustered at plant level) with 500 replications. *** p<0.01, **
p<0.05, * p<0.1
Page 21
AC
CE
PT
E
D
MA
NU
SC
RI
PT
ACCEPTED MANUSCRIPT
Page 22
ACCEPTED MANUSCRIPT
Table A3: Effects of Plant Divestiture: Excluding reactors that were permanently shutdown
Using plant-centered events counted from the NRC spread file, and normalizing by Critical Year
Poisson
(3)
Linear Panel
(4)
Panel A: for the period of 1988-2006
-0.151
Divestiture
(0.118)
-0.148
(0.117)
-0.145
(0.119)
-0.141
(0.199)
Panel B: for the period of 1988-2007
-0.136
Divestiture
(0.119)
-0.137
(0.118)
-0.131
(0.120)
-0.122
(0.199)
Panel C: for the period of 1988-2008
-0.250**
Divestiture
(0.119)
-0.250**
(0.119)
-0.244**
(0.120)
-0.194
(0.197)
Panel D: for the period of 1988-2009
-0.288**
Divestiture
(0.124)
-0.289**
(0.127)
Panel E: for the period of 1988-2010
-0.255**
Divestiture
(0.123)
Panel F: for the period of 1988-2011
-0.236*
Divestiture
(0.122)
CE
Panel I: for the period of 1988-2014
-0.213*
Divestiture
(0.121)
SC
-0.282**
(0.125)
-0.215
(0.190)
-0.251**
(0.125)
-0.249**
(0.124)
-0.191
(0.190)
-0.233*
(0.125)
-0.230*
(0.122)
-0.167
(0.188)
-0.220*
(0.124)
-0.219*
(0.122)
-0.156
(0.189)
-0.223*
(0.123)
-0.224*
(0.123)
-0.156
(0.191)
-0.210*
(0.121)
-0.208*
(0.121)
-0.142
(0.190)
-0.102
(0.122)
-0.0785
(0.195)
MA
PT
E
Panel H: for the period of 1988-2013
-0.230*
Divestiture
(0.123)
D
Panel G: for the period of 1988-2012
-0.224*
Divestiture
(0.122)
RI
PT
Conditional NB
(2)
NU
Negative Binomial
(1)
AC
Panel J: for the period of 1988-2014 but excluding 2008 and 2009
-0.107
-0.103
Divestiture
(0.122)
(0.120)
Note: We run unconditional negative binomial, conditional negative binomial, Poisson, and linear panel regressions, as in
Hausman (2014). The dependent variable in all regressions is the number of “plant-centered” initiating events counted in
the NRC spreadsheet, by reactor-year. We use reactor critical year as the normalizing variable and vary the study period in
these regressions. We exclude San Onofre (CA), Crystal River (FL), Kewaunee (WI) and Vermont Yankee (VT), which
were shutdown permanently between 2013 and 2014, either due to technical issue or financial consideration. Following
Hausman (2014), all regressions include reactor fixed effects and year dummies, with robust standard errors clustered at
plant level in Negative Binomial, Poisson, and Panel Linear regressions; for Conditional NB regressions, standard errors
are bootstrapped (and also clustered at plant level) with 500 replications. *** p<0.01, ** p<0.05, * p<0.1
Page 23
ACCEPTED MANUSCRIPT
Table A4: Effects of Plant Divestiture: Including power uprates as independent variables
Using plant-centered events counted from the NRC spread file, and normalizing by Critical Year
Conditional NB
(2)
Poisson
(3)
Linear Panel
(4)
Panel A: for the period of 1988-2006
-0.154
Divestiture
(0.120)
-0.154
(0.124)
-0.150
(0.120)
-0.174
(0.197)
0.0547
(0.101)
0.0552
(0.107)
0.0507
(0.101)
0.0451
(0.242)
0.172
(0.166)
Panel B: for the period of 1988-2007
-0.135
Divestiture
(0.125)
0.186
(0.183)
0.0554
(0.102)
0.0562
(0.109)
SPU
RI
PT
Negative Binomial
(1)
0.195
(0.171)
0.169
(0.267)
-0.132
(0.125)
-0.154
(0.199)
0.0515
(0.102)
0.0511
(0.239)
0.186
(0.166)
0.168
(0.253)
-0.255**
(0.129)
-0.247**
(0.123)
-0.226
(0.196)
0.0306
(0.106)
0.0254
(0.0996)
0.0279
(0.233)
0.123
(0.159)
0.132
(0.153)
0.130
(0.228)
-0.304**
(0.136)
-0.294**
(0.127)
-0.257
(0.191)
0.0427
(0.0982)
0.0411
(0.103)
0.0387
(0.0979)
0.0420
(0.224)
0.119
(0.136)
Panel E: for the period of 1988-2010
-0.261**
Divestiture
(0.125)
0.136
(0.146)
0.138
(0.142)
0.144
(0.211)
-0.260*
(0.134)
-0.256**
(0.126)
-0.226
(0.190)
0.0285
(0.100)
0.0247
(0.103)
0.0247
(0.0994)
0.0232
(0.222)
0.0446
(0.133)
Panel F: for the period of 1988-2011
-0.242*
Divestiture
(0.124)
0.0589
(0.139)
0.0632
(0.139)
0.0961
(0.204)
-0.243*
(0.132)
-0.238*
(0.125)
-0.205
(0.189)
SPU
0.0175
(0.0992)
0.0185
(0.103)
0.0143
(0.0985)
0.00933
(0.220)
EPU
0.0443
(0.129)
0.0611
(0.137)
0.0615
(0.135)
0.115
(0.199)
EPU
0.166
(0.161)
Panel C: for the period of 1988-2008
-0.251**
Divestiture
(0.123)
0.178
(0.170)
NU
EPU
0.112
(0.148)
Panel D: for the period of 1988-2009
-0.299**
Divestiture
(0.127)
MA
0.0290
(0.0995)
SPU
EPU
PT
E
AC
CE
EPU
SPU
D
EPU
SPU
SC
SPU
-0.139
(0.130)
Page 24
Negative Binomial
(1)
Poisson
(3)
Linear Panel
(4)
Panel G: for the period of 1988-2012
-0.235*
Divestiture
(0.125)
-0.236*
(0.133)
-0.232*
(0.125)
-0.200
(0.190)
0.00709
(0.0994)
0.00471
(0.102)
0.00452
(0.0987)
0.000607
(0.219)
0.0687
(0.130)
Panel H: for the period of 1988-2013
-0.256**
Divestiture
(0.127)
0.0871
(0.140)
0.0833
(0.134)
0.151
(0.192)
-0.254*
(0.135)
-0.252**
(0.128)
-0.211
(0.191)
-0.0105
(0.103)
-0.0123
(0.106)
-0.0135
(0.103)
-0.00940
(0.221)
0.197
(0.136)
Panel I: for the period of 1988-2014
-0.238*
Divestiture
(0.126)
0.206
(0.148)
0.209
(0.139)
0.257
(0.193)
-0.240*
(0.133)
-0.234*
(0.126)
-0.195
(0.191)
-0.0198
(0.103)
-0.0214
(0.108)
-0.0224
(0.103)
-0.0182
(0.220)
0.185
(0.134)
0.242
(0.185)
-0.123
(0.128)
-0.125
(0.195)
-0.0220
(0.105)
-0.0227
(0.227)
SPU
EPU
SPU
MA
SPU
NU
EPU
SC
Conditional NB
(2)
RI
PT
ACCEPTED MANUSCRIPT
0.173
0.184
(0.132)
(0.143)
Panel J: for the period of 1988-2014 but excluding 2008 and 2009
-0.125
-0.125
Divestiture
(0.128)
(0.132)
PT
E
D
EPU
-0.0200
(0.106)
SPU
-0.0189
(0.111)
0.196
0.204
0.208
0.257
(0.147)
(0.158)
(0.149)
(0.199)
Note: We run unconditional negative binomial, conditional negative binomial, Poisson, and linear panel regressions, as in
Hausman (2014). The dependent variable in all regressions is the number of “plant-centered” initiating events counted in
the NRC spreadsheet, by reactor-year. We use reactor critical year as the normalizing variable and vary the study period in
these regressions. We include two binary indicators for major power uprates
and
, indicating whether a reactor
has implemented a Stretch Power Uprate or Extended Power Uprate. Following Hausman (2014), all regressions include
reactor fixed effects and year dummies, with robust standard errors clustered at plant level in Negative Binomial, Poisson,
and Panel Linear regressions; for Conditional NB regressions, standard errors are bootstrapped (and also clustered at plant
level) with 500 replications. *** p<0.01, ** p<0.05, * p<0.1
AC
CE
EPU
Page 25
ACCEPTED MANUSCRIPT
Table A5: Effects of Plant Divestiture:
Replicating Hausman (2014) and Only Varying the Study Period
Negative Binomial
(1)
Conditional NB
(2)
Poisson
(3)
Linear Panel
(4)
Panel A: Using Hausman (2014) Data and Normalizing by Capacity Factor, for the period of 1988-2006
-0.216*
(0.124)
-0.21*
(0.11)
-0.20
(0.12)
RI
PT
Divestiture
-0.49**
(0.23)
Panel B: Using Hausman (2014) Data and Normalizing by Capacity Factor, for the period of 1988-2007
-0.194
(0.123)
Divestiture
-0.20*
(0.11)
-0.18
(0.12)
-0.46*
(0.23)
-0.31***
(0.10)
NU
-0.301**
(0.120)
Divestiture
SC
Panel C: Using Hausman (2014) Data and Normalizing by Capacity Factor, for the period of 1988-2008
-0.28**
(0.12)
-0.53**
(0.22)
Panel D: Using Hausman (2014) Data and Normalizing by Capacity Factor, for the period of 1988-2009
-0.32**
(0.12)
-0.54**
(0.23)
X
X
X
X
X
X
X
X
D
Reactor fixed effects
Year dummies
-0.34***
(0.12)
MA
-0.335***
(0.122)
Divestiture
AC
CE
PT
E
Note: We replicate Hausman (2014) with the data from Hausman (2014) and using capacity factors for normalization, only
varying the study period. We run unconditional negative binomial, conditional negative binomial, Poisson, and linear
panel regressions as in Hausman (2014). Robust standard errors are clustered at plant level in Negative Binomial, Poisson,
and Panel Linear regressions; for Conditional NB regressions, standard errors are bootstrapped (and also clustered at plant
level) with 500 replications. *** p<0.01, ** p<0.05, * p<0.1
Page 26
ACCEPTED MANUSCRIPT
Online Appendix B:
AC
CE
PT
E
D
MA
NU
SC
RI
PT
Initiating Events Excluded in Our Study
Page 27
ACCEPTED MANUSCRIPT
Arkansas 1
3131998005
1998-12-28
Arkansas 1
Arkansas 2
3132003001
3682014002
2003-08-29
2014-04-03
Beaver Valley 1
3341993013
1993-10-12
Beaver Valley 1
Beaver Valley 2
3341997005
3341994005
1997-03-19
1994-06-01
Beaver Valley 2
3341997005
1997-03-19
Braidwood 1
Braidwood 1
4561988022
4561990008
1988-10-16
1990-06-08
Braidwood 1
4562007001
2007-06-27
Braidwood 2
Braidwood 2
4571989002
4571989004
1989-05-11
1989-09-07
Browns Ferry 2
2592011001
2011-04-27
Browns Ferry 3
Browns Ferry 3
2592011001
2962004002
2011-04-27
2004-11-23
Browns Ferry 3
2962005001
2005-02-11
Browns Ferry 3
Brunswick 1
2962005003
3251995011
2005-10-31
1995-05-19
Brunswick 1
3252004002
2004-08-14
Brunswick 2
Byron 1
3241999006
4541988005
1999-06-28
1988-08-04
Byron 1
4541990011
1990-08-19
Byron 2
Byron 2
4551987019
4552000001
1987-10-02
2000-01-13
Callaway
4831995004
1995-06-08
Callaway
Callaway
4832000002
4832004003
2000-02-13
2004-02-03
Calvert Cliffs 1
3171993003
1993-06-10
D
PT
E
CE
AC
Calvert Cliffs 1
Calvert Cliffs 1
3171999004
3172010001
1999-07-24
2010-02-18
3172010003
2010-05-12
Calvert Cliffs 1
Calvert Cliffs 2
3172011001
3181996001
2011-08-27
1996-02-27
Catawba 1
4132006001
2006-05-20
Catawba 2
Clinton 1
4132006001
4611996004
2006-05-20
1996-04-09
Clinton 1
4612004003
2004-07-13
Comanche Peak 1
4451996007
1996-08-09
Calvert Cliffs 1
Comanche Peak 2
4452003003
2003-05-15
Comanche Peak 2
Cook 1
4461996006
3151996004
1996-09-18
1996-09-22
Cooper
2982003006
2003-10-28
Crystal River 3
Crystal River 3
3021989023
3021991003
1989-06-16
1991-04-20
Crystal River 3
3021996017
1996-05-31
Crystal River 3
Crystal River 3
3022002002
3022004003
2002-11-07
2004-09-06
Davis-Besse
RI
PT
Event Date
1994-04-11
1998-12-25
Event Date
1997-10-27
2003-05-15
3461998006
1998-06-24
Diablo Canyon 1
Diablo Canyon 1
2751990005
2751994020
1990-06-14
1994-12-14
Diablo Canyon 1
2751995017
1995-12-13
Diablo Canyon 1
Diablo Canyon 1
2751996012
2751999006
1996-08-10
1999-09-22
Diablo Canyon 1
2751999009
1999-10-28
Diablo Canyon 2
Diablo Canyon 2
2751994020
2751996012
1994-12-14
1996-08-10
Diablo Canyon 2
2751999009
1999-10-28
Diablo Canyon 2
Diablo Canyon 2
3231994012
3231995002
1994-12-19
1995-09-23
Diablo Canyon 2
3231998005
1998-12-01
Diablo Canyon 2
Diablo Canyon 2
3232008002
3232013005
2008-10-21
2013-07-10
Dresden 2
2372000004
2000-11-30
Dresden 3
Dresden 3
2491989001
2491995019
1989-03-25
1995-10-29
Dresden 3
2492004003
2004-05-05
Farley 1
Farley 1
3481991009
3482008004
1991-08-19
2008-11-19
Farley 2
3641991005
1991-08-06
Fermi 2
Fermi 2
3411988019
3411998001
1988-05-07
1998-02-01
Fermi 2
3412003002
2003-08-14
Fermi 2
FitzPatrick
3412010002
3331990023
2010-06-06
1990-10-19
FitzPatrick
3331993004
1993-02-25
FitzPatrick
FitzPatrick
3332003001
3332007002
2003-08-14
2007-09-12
Fort Calhoun
2851993011
1993-06-24
SC
LER
3131994002
3131998005
LER
4451997009
4452003003
MA
Plant Name
Arkansas 1
Arkansas 1
Plant Name
Comanche Peak 1
Comanche Peak 1
NU
Table B.1 Initiating Events with external root
causes
ACCEPTED MANUSCRIPT
LER
2441992003
2442003002
Event Date
1992-02-29
2003-08-14
Plant Name
La Salle 1
La Salle 2
LER
3732013002
3732013002
Event Date
2013-04-17
2013-04-17
2442003005
2003-10-15
Limerick 1
3521995002
1995-02-21
Grand Gulf
Grand Gulf
4161989010
4161989016
1989-07-22
1989-11-07
Limerick 2
Limerick 2
3521995002
3531996004
1995-02-21
1996-05-14
Grand Gulf
4161991005
1991-06-17
Limerick 2
3532004001
2004-06-22
Grand Gulf
Grand Gulf
4161991010
4161991012
1991-08-10
1991-11-19
McGuire 1
McGuire 2
3691991001
3701993008
1991-02-11
1993-12-27
Grand Gulf
4161992010
1992-06-06
Millstone 2
3362008003
2008-05-22
Grand Gulf
Grand Gulf
4162000005
4162001003
2000-09-15
2001-08-07
Millstone 2
Millstone 2
3362009001
3362014006
2009-07-03
2014-05-25
Grand Gulf
4162003002
2003-04-24
Millstone 3
3362014006
2014-05-25
Harris
Harris
4001996008
4002002003
1996-04-25
2002-08-15
Millstone 3
Millstone 3
4231988024
4231989008
1988-10-22
1989-05-06
Harris
4002003005
2003-08-17
Millstone 3
4231990011
1990-03-30
Hatch 1
Hatch 1
3211991001
3211992021
1991-01-18
1992-08-27
Millstone 3
Millstone 3
4231991014
4231992027
1991-06-09
1992-11-05
Hope Creek
3541990003
1990-03-19
Millstone 3
4231998044
1998-11-11
Hope Creek
Indian Point 2
3542003007
2471995016
2003-09-19
1995-06-12
Millstone 3
Monticello
4232005003
2631991019
2005-09-29
1991-08-25
Indian Point 2
2471996003
1996-03-05
Monticello
2631994003
1994-04-15
Indian Point 2
Indian Point 2
2471996016
2471997002
1996-08-22
1997-01-26
Monticello
Nine Mile Pt. 1
2631994004
2201987024
1994-06-04
1987-12-07
Indian Point 2
2471997018
1997-07-26
Nine Mile Pt. 1
2201994002
1994-04-05
Indian Point 2
Indian Point 2
2471999015
2472001007
1999-08-31
2001-12-26
Nine Mile Pt. 1
Nine Mile Pt. 2
2202003002
4102003002
2003-08-14
2003-08-14
Indian Point 2
2472003003
2003-04-28
North Anna 1
3381988005
1988-01-13
Indian Point 2
Indian Point 2
2472003004
2472003005
2003-08-03
2003-08-14
North Anna 1
North Anna 2
3382011003
3382011003
2011-08-23
2011-08-23
Indian Point 3
2861990004
1990-06-29
North Anna 2
3391998004
1998-09-17
Indian Point 3
Indian Point 3
2861991004
2862000008
1991-03-20
2000-06-09
North Anna 2
North Anna 2
3392005001
3392010002
2005-08-05
2010-05-28
Indian Point 3
2862002003
2002-11-15
North Anna 2
3392010004
2010-06-16
2862003003
2862003005
2003-06-22
2003-08-14
Oconee 1
Oconee 2
2692007001
2692007001
2007-02-15
2007-02-15
3051992017
1992-09-15
Oconee 2
2701992004
1992-10-19
Kewaunee
La Salle 1
3051992020
3731989009
1992-11-01
1989-03-02
Oconee 2
Oconee 2
2701995002
2701997002
1995-04-14
1997-07-06
La Salle 1
3731990006
1990-03-28
Oyster Creek
2191992005
1992-05-03
La Salle 1
La Salle 1
3731995016
3732001001
1995-09-24
2001-01-31
Oyster Creek
Oyster Creek
2191994007
2191997010
1994-05-31
1997-08-01
La Salle 1
3732011001
2011-02-01
Oyster Creek
2192003003
2003-08-14
Kewaunee
SC
NU
MA
PT
E
CE
AC
Indian Point 3
Indian Point 3
RI
PT
Ginna
D
Plant Name
Ginna
Ginna
Page 29
ACCEPTED MANUSCRIPT
Plant Name
Oyster Creek
Oyster Creek
LER
2192005002
2192009005
Event Date
2005-06-01
2009-07-12
Plant Name
River Bend
River Bend
LER
4581989042
4581992005
Event Date
1989-12-01
1992-03-05
2192012001
2012-07-23
River Bend
4581999014
1999-10-29
Palisades
Palo Verde 1
2552002002
5281991010
2002-12-01
1991-10-27
River Bend
River Bend
4582004001
4582004002
2004-08-15
2004-10-01
Palo Verde 1
5281996004
1996-08-10
River Bend
4582007002
2007-05-04
Palo Verde 1
Palo Verde 2
5282004006
5282004006
2004-06-14
2004-06-14
Salem 1
Salem 1
2721991024
2722003002
1991-06-16
2003-07-29
Palo Verde 2
5291997006
1997-10-20
Salem 1
2722007002
2007-04-24
Palo Verde 2
Palo Verde 3
5292004002
5281991010
2004-07-14
1991-10-27
Salem 1
Salem 1
2722007002
2722011003
2007-04-30
2011-04-21
Palo Verde 3
5281996004
1996-08-10
Salem 2
3111991017
1991-11-09
Palo Verde 3
Palo Verde 3
5282004006
5301991008
2004-06-14
1991-11-14
Salem 2
San Onofre 2
3112010001
3612011002
2010-01-03
2011-09-08
5302003004
2003-07-28
San Onofre 3
3612011002
2011-09-08
Peach Bottom 2
Peach Bottom 2
2771992012
2771992015
1992-07-17
1992-08-17
San Onofre 3
San Onofre 3
1993-01-16
2002-02-27
Peach Bottom 2
2772003003
2003-07-22
Seabrook
4431991008
1991-06-27
Peach Bottom 2
Peach Bottom 3
2772003004
2772003004
2003-09-15
2003-09-15
NU
3621993001
3622002001
Seabrook
Seabrook
4431993009
4431998014
1993-05-20
1998-12-22
Peach Bottom 3
2781991010
1991-07-07
Seabrook
4432008001
2008-01-19
Peach Bottom 3
Perry
2781992003
4402003002
1992-05-04
2003-08-14
Sequoyah 1
Sequoyah 1
3271992018
3271996006
1992-10-26
1996-06-23
Pilgrim
2931990008
1990-05-13
Sequoyah 2
3281995007
1995-12-21
Pilgrim
Pilgrim
2931992016
2931993004
1992-12-13
1993-03-13
South Texas 1
South Texas 1
4981988026
4981990014
1988-03-30
1990-06-20
Pilgrim
2931993022
1993-09-10
South Texas 1
4981995013
1995-12-18
Pilgrim
Pilgrim
2931994005
2932008006
1994-08-29
2008-12-19
South Texas 2
South Texas 2
4991992003
4992001002
1992-02-24
2001-03-01
Pilgrim
2932013003
2013-02-08
South Texas 2
4992013002
2013-01-08
Pilgrim
Point Beach 1
2932013009
2661995006
2013-10-14
1995-07-14
St. Lucie 1
St. Lucie 1
3351993007
3351993007
1993-09-18
1993-09-20
Point Beach 1
2662000001
2000-01-21
St. Lucie 1
3351993007
1993-09-22
2662000010
3012004002
2000-10-27
2004-05-05
St. Lucie 1
St. Lucie 1
3351994005
3351994007
1994-06-06
1994-10-26
2821996012
1996-06-29
St. Lucie 1
3352011001
2011-08-22
Prairie Island 1
Prairie Island 2
2821997008
2821996012
1997-06-02
1996-06-29
St. Lucie 2
Surry 1
3892006001
2802011001
2006-01-20
2011-04-16
Prairie Island 2
3062011002
2011-05-09
Surry 2
2802011001
2011-04-16
Quad Cities 1
Quad Cities 2
2541990004
2651993024
1990-03-10
1993-12-02
Surry 2
Susquehanna 1
2812004001
3871988010
2004-05-21
1988-06-18
Quad Cities 2
2652001001
2001-08-02
Susquehanna 1
3871989027
1989-12-24
Prairie Island 1
PT
E
CE
AC
Point Beach 1
Point Beach 2
MA
Palo Verde 3
D
SC
RI
PT
Oyster Creek
Page 30
Susquehanna 2
3881991012
1991-08-06
Susquehanna 2
Susquehanna 2
3881995005
3882005005
1995-04-15
2005-06-06
Three Mile Isl 1
2891997007
1997-06-21
Three Mile Isl 1
Turkey Point 3
2892006003
2501995007
2006-12-13
1995-10-17
Turkey Point 3
2502008001
2008-02-26
Turkey Point 4
Vermont Yankee
2502008001
2711991005
2008-02-26
1991-03-13
Vermont Yankee
2711991009
1991-04-23
Vermont Yankee
Vermont Yankee
2711991014
2711998016
1991-06-15
1998-06-09
Vermont Yankee
2712005001
2005-07-25
Vermont Yankee
Vogtle 1
2712010001
4241988025
2010-05-26
1988-07-31
Vogtle 1
4241995002
1995-07-23
Vogtle 2
Vogtle 2
4241995002
4251994001
1995-07-23
1994-01-07
Vogtle 2
4251998003
1998-05-09
Waterford 3
Waterford 3
3821990003
3821990012
1990-03-29
1990-08-25
Waterford 3
3821995002
1995-06-10
Wolf Creek
Wolf Creek
4821992016
4821996002
1992-11-10
1996-01-30
Wolf Creek
4822000003
2000-09-04
Wolf Creek
Wolf Creek
4822004005
4822009002
2004-10-07
2009-08-19
Wolf Creek
4822012001
SC
Event Date
1991-07-31
1990-02-06
NU
LER
3871991008
3881990002
2012-01-13
AC
CE
PT
E
D
MA
Plant Name
Susquehanna 1
Susquehanna 2
RI
PT
ACCEPTED MANUSCRIPT
Page 31
ACCEPTED MANUSCRIPT
Calvert Cliffs 2
3181988004
1988-04-27
Calvert Cliffs 2
Clinton 1
3182013003
4611993006
2013-05-08
1993-12-06
Clinton 1
4612013008
2013-12-08
Comanche Peak 1
Comanche Peak 1
4451991020
4452010001
1991-07-13
2010-01-09
Comanche Peak 2
4461996007
1996-10-18
Comanche Peak 2
Cook 2
4461999002
3161990004
1999-01-03
1990-06-11
Diablo Canyon 2
3231997003
1997-07-02
Dresden 2
Dresden 2
2371990002
2371995009
1990-01-16
1995-03-05
Farley 2
3642003001
2003-11-10
Grand Gulf
Hatch 1
4162007003
3211990013
2007-08-21
1990-06-20
Hatch 1
3211993013
1993-10-22
Hatch 2
Indian Point 3
3662001002
2862014004
2001-10-26
2014-08-13
La Salle 1
3731991006
1991-05-19
La Salle 2
Limerick 2
3741990010
3531993001
1990-09-12
1993-01-03
McGuire 1
3691990001
1990-01-08
McGuire 2
McGuire 2
3701987021
3701989001
1987-11-30
1989-03-03
Millstone 2
3361991004
1991-02-16
Millstone 3
Monticello
4232013007
2631991014
2013-08-09
1991-06-05
North Anna 1
3381988002
1988-01-08
D
PT
E
CE
AC
Oconee 1
Oconee 2
2691990013
2701989003
1990-08-28
1989-02-05
2871988006
1988-11-14
Oconee 3
Oconee 3
2871989002
2871990001
1989-03-06
1990-01-19
Palisades
2551990001
1990-01-09
Palisades
Palo Verde 1
2552008003
5281995008
2008-05-23
1995-05-30
Pilgrim
2931993014
1993-05-31
Point Beach 1
2661991005
1991-05-30
Oconee 3
LER
2662015002
3012001001
Event Date
2014-12-02
2001-02-06
Prairie Island 2
3061998005
1998-11-09
Quad Cities 1
Robinson 2
2541988016
2611996007
1988-12-05
1996-10-20
Seabrook
4431990022
1990-08-22
South Texas 1
Vogtle 1
4981990006
4241989012
1990-07-30
1989-05-09
Waterford 3
3821994007
1994-04-26
RI
PT
Event Date
2010-05-05
1988-06-02
SC
LER
3252010003
4551988006
MA
Plant Name
Brunswick 1
Byron 2
Plant Name
Point Beach 1
Point Beach 2
NU
Table B.2 Initiating Events with indeterminate root
causes
Документ
Категория
Без категории
Просмотров
0
Размер файла
308 Кб
Теги
015, eneco, 2017
1/--страниц
Пожаловаться на содержимое документа